The output signal of photon counting lidar has serious distortion phenomenon under the high-flux condition,which affects the extraction of target information.Therefore,to investigate the change rule of signal distortion,this paper establishes the steady-state forward recursion model for signal calculation of photon counting lidar under the condition of multi-trigger.The model has many advantages,such as high precision,fast speed and accurate reflection of physical process.The influence of the echo signal distortion is explored by changing the signal amplitude N,the pulse width PW,the background noise Nb and dead time Td of the model.Results show that when the dead-pulse-ratio Rdp≤2,the signal distortion is related to N,PW and Td,when Rdp>2,the signal distortion is only related to N.At the same time, our simulations show that when the detection loss rate α≤5%,the influence of signal distortion can be ignored, which can be used to distinguish high-flux signal distortion (α>5%) and the fidelity of the low-flux signal (α≤5%).According to the calculation,changing the length of the dead time converts the high-flux to the low-flux under some specific conditions.Compared with the traditional "5%" criterion based on photon count rate,the above conclusion provides a more accurate boundary between the high-and low-flux and signal distortion.

For large ring laser gyroscope (RLG),lasers require more accurate and more stable excitation power supplies.Aiming at the poor output performance of single-ended flyback converter,a single-ended flyback converter with capacitor-inductor-capacitor (CLC) filter is designed by adding capacitors and inductor to the output end of the traditional flyback converter.On this basis,the static working point of the converter is determined by using the switch elements average model method,and the fourth-order small signal model operating in continuous conduction mode (CCM) is established.The correlation transfer function is derived and the phase leading correction is introduced into the system to improve its stability.The system model of proportion-integral-differential (PID) control is established via MATLAB,and the analysis and verification of flyback converter with CLC filter is completed.The results show that the model established by small signal modeling theory is reasonable and available,which is helpful for the design and optimization of laser excitation power supply.

In order to solve the problem that the existing algorithms pay less attention to the interactive selection between features from different sources and the extraction of cross modal features is insufficient,a RGB-D visual saliency detection network based on extracting bi-directional selection dense features is proposed.First,in order to filter out the features that can enhance the saliency areas of RGB images and depth images at the same time,a bi-directional selection module (BSM) is introduced. In order to solve the problem of insufficient cross modal feature extraction,which leads to redundant calculation and low accuracy,a dense extraction module (DEM) is introduced.Finally,the dense features are cascaded and fused through the feature aggregation module (FAM),and the recurrent residual refinement aggregating module (RAM) is combined with the deep supervision to achieve the continuous optimization of the coarse saliency maps,and finally the accurate saliency maps are obtained.Comprehensive experiments on four widely used datasets show that the proposed algorithm is superior to seven existing methods in four key indicators.

An image enhancement algorithm based on unsupervised learning is proposed,aim at the problems of low illumination image enhancement algorithms,suffering from loss of recovery details,high network complexity,and difficulty obtaining paired data sets.In YIQ color space,the enhancement curve of luminance channel Y is calculated by the constructed lightweight network and power index function to get the image of the enhancement of the poorly exposed area and the containment of the high light area.The no-reference loss function used in this network can implicitly evaluate image enhancement quality and drive network learning.Experimental results show that the proposed algorithm achieves competitive results regarding visual effects and image quality when the trainable parameters and model weight only account for 9.5 k/88 kB.

The traditional sparrow search algorithm (SSA) has the problems that it is easy to fall into the local optimum and the search ability is insufficient in the process of optimization.In order to solve the above problems,an improved sparrow search algorithm (ISSA) based on Gaussian cloud improvement is proposed.First,Bernoulli chaotic mapping is used to initialize the population to improve the initial population quality of the algorithm;secondly,an adaptive Gaussian cloud mutation strategy is introduced in the update of the finder position to improve the global search ability of the algorithm in the iterative process;finally,the reverse t distribution learning strategy is used to perform random reverse learning on the optimal position to improve the algorithm′s ability to jump out of the local optimum.In the simulation experiment,this algorithm is compared with other four basic algorithms with 13 benchmark functions,and compared with other ISSAs.The experimental results show that the algorithm has good convergence and accuracy,and the global exploration ability is greatly improved compared with the original algorithm.The ISSA is applied to the Kapur entropy multi-threshold image segmentation task,and the results show that ISSA has higher segmentation accuracy than the other four basic algorithms.

Aiming at problems of halo phenomenon,color distortion of bright areas and inaccurate estimation of ambient light generated by dark channel prior (DCP),a single image defogging algorithm based on superpixel dark channel and auto-color optimization is proposed in this paper.First,the improved White Patch Retinex algorithm is used to enhance the image and calculate the accurate environmental light.Then,the robustness and accuracy of transmission estimation is improved by using superpixel image segmentation and guided filtering algorithm in traditional dark channel defogging algorithm,and adaptive tolerance is used to compensate the transmittance of bright region,which can effectively suppress the color distortion in bright region.Finally,an auto-color optimization algorithm is used to improve image contrast.Comparative experiments of different algorithms are carried out from both subject and object dimensions.Experimental results show that the entropy increases by 0.2 bit,the peak signal-to-noise ratio (PSNR) increases by about 0.8 dB,and the running efficiency is increased using different algorithms for natural fog image with different concentrations.The algorithm has good adaptability to different fog images with different concentrations in different scenes,and the restored images have true colors,clear texture and rich details Defogging effect is good.

Accurate assessment of the metallographic grain size grade of steel can detect material deterioration and ensure the safety of equipment in service.In order to solve the problems that the traditional manual evaluation of steel metallographic grain size grade is time-consuming and easily influenced by manual experience,the evaluation results are not consistent and irreducible,etc,a deep learning-based steel metallographic grain size grade evaluation method is proposed.By adding a jump connection layer to the U-net model and reducing the number of downsampling to improve the segmentation accuracy and reduce the number of network parameters,the pixel accuracy is 93.86% and the mean pixel accuracy (MPA) is 86.89% on the 117 validation sets.The number of network parameters is only 2.02 M.The grain boundary prediction results are digitally processed and combined with the intercept point method to grade the grain size,and the average time taken to grade the grain size of steel on the test image is only 8.3 s/sheet.Compared with manual rating methods,this method is accurate,efficient and repeatable.

Large geodesic laser gyroscope is an inertial sensor based on Sagnac effect,which can accurately monitor the angular velocity of the earth rotation.It has a good application prospect in the fields of universal time,seismic wave detection and fundamental physics.The geodesic laser gyroscope is sensitive to the change of ambient temperature.The change of temperature changes the cavity length of the geodesic laser gyroscope,and then causes the change of the scale factor,and finally affects the test accuracy of the geodesic laser gyroscope.In this paper,the relative error of geodesic laser gyroscope is analyzed and discussed by measuring the temperature in different working environments when the temperature changes and the heat source location is different.When the temperature change is maximum,the relationship between the material usage and the relative error of the ring cavity of the laser gyroscope is analyzed.The results show that the working environment of Pucheng Laboratory is better than that of Lintong Laboratory when the temperature change in the laboratory is the largest,and the relative error of different configurations of geodesic laser gyro ranges from 10-6 to 10-9.The relative error decreases when the amount of Zerodur is increased in geodesic laser gyroscope.

In order to improve the utilization rate of optical signals in radio over fiber (RoF) system,a RoF system based on two paralleled Mach-Zehnder modulators (MZM) is proposed.The two paralleled MZMs are modulated by radio frequency (RF) signal and five optical sideband signals can be generated,which are ±1st order sidebands,±2nd order sidebands and optical carrier.±2nd order optical sidebands are modulated by baseband data,millimeter (mm)-wave signals with modulated data can be generated after beating by photodiode (PD),which can be emitted by aerials.+2nd order sideband and optical carrier are used to generate the mm-wave signals without modulated data after beating by PD,this mm-wave signals can be used as the local oscillator (LO) for the demodulation of receiving end.-2nd order optical sideband is used as the optical carrier for uplink.In the proposed scheme,the whole five optical sideband signals are fully utilized,the utilization rate of optical signals is improved.In addition,the analytical models for transmission through a dispersive medium are confirmed.Results show that such a scheme can offer a realistic solution for RoF system.

Image super-resolution is widely used in medical and security fields.Aiming at the shortcomings that traditional super-resolution reconstruction (SR) methods cannot reconstruct edge feature images,this paper proposes a reconstruction scheme based on prior information and dense connected network model.By taking into account the different combinations of residual features of input statistical information,a multi attention module is introduced to improve the network performance without adding additional modules by cooperating with the backbone network structure.The proposed model has better performance than the existing state of the art (SOTA) model with complex structures. In order to avoid the sharp drift of the input identity features,a network module of attention mechanism based on prior information is proposed to estimate the real low resolution (LR) counterpart.This model has advantages in terms of capturing motion noise,etc.The experimental results show that this method has more advantages in evaluation indicators and subjective visual analysis than other mainstream methods.

The continuous flash suppression paradigm (CFS) is currently the most up-to-date subconscious visual perception research paradigm.However,due to the lack of mature visual experiment platform,the progress of CFS-related research is seriously restricted.In view of this,our team have designed a CFS-based visualization experiment platform including a hardware system developed with standardized precision optical components and a software system developed with Psychtoolbox under MATLAB.Among them,the experimental software system not only provides a fully functional and convenient visualization setting interface for experimental parameters,but also creates an automatic program for the generation of Mondrian masking images with different physical features.In order to evaluate the performance of the experimental platform, we have carried out a subconscious visual perception experiment with picture and video stimuli.The results show that the experimental platform can provide convenient experimental settings,good masking effect,stable program running,and easy data access.It can better serve the design and development of psychophysical experiments related to CFS and effectively improve the level of unbconscious visual cognition research.

The corona virus disease 2019 (COVID-19) is severely affects the development of society and economy,and threatens human health.In order to solve the problem that how to identify patients infected with the virus more accurately and quickly,convolutional neutral network (CNN) methods are used to identify COVID-19 chest X-ray images.However,due to the low recognition accuracy of CNN,it is difficult to accurately determine whether a patient is infected with COVID-19.In order to improve the recognition performance of the network for COVID-19 chest X-ray images,firstly,the attention steered trapezoid pyramid fusion network (ASTPNet) is proposed.The ASTPNet can be attached to different CNNs.The characteristics of deep and shallow networks in the model are effectively utilized.Secondly,the attention steered block (AS Block) is proposed to aggregate the weighted information more efficiently to emphasize effective semantic information in channels and spaces,and weaken ineffective interference information through channel and spatial attention.The results show that the accuracy is significantly improved after attaching the ASTPNet to VGG16/19,ResNet34/50 and ResNeXt.When applied to the self-built COVID-19 dataset,and compared with other CNN methods,ASTP-ResNet34 has the better performance.The accuracy reaches 98.40% (two classes) and 97.10% (three classes).It can accurately determine whether the infection of COVID-19.